Claude Certified Architect — Foundations (CCAR-F)
Anthropic's foundational certification on designing and deploying agentic applications with Claude at production scale.
Unofficial material, not affiliated with Anthropic. Questions are AI-generated based on public documentation, with no guarantee of passing the official exam.
Domains covered
Quick facts
- Official name
- Claude Certified Architect — Foundations (CCA-F)
- Format
- 60 multiple-choice questions, across 4 scenarios drawn from a pool of 6
- Duration
- 120 minutes, single sitting, closed-book
- Passing score
- 720, on a 100–1,000 scale
- Official exam fee
- US$125 per attempt (confirm on the official source)
- Certificate validity
- 12 months from the issue date
Start with the Study Track
21 free lessons, each with a checkpoint quiz, covering the same domains as the exam — before you move on to the full mock exam.
Start the free trackFrequently asked questions
Is there a fee for the official CCA-F exam?
Yes. Anthropic charges a US$125 registration fee per attempt through the Partner Academy (Skilljar). AICert Study's mock exams are separate from that fee and are meant for practice before you book the official exam.
How many questions does the exam have, and how much time do I get?
60 multiple-choice questions, organized into 4 scenarios drawn from a pool of 6, to answer in a single 120-minute sitting.
Does the CCA-F certificate expire?
Yes, the credential is valid for 12 months from its issue date, per Anthropic's official Exam Guide.
Are AICert Study's mock exams made by Anthropic?
No. AICert Study is an independent platform with no affiliation to Anthropic. Questions are AI-generated from public documentation for practice purposes — they don't replace or guarantee a pass on the official exam.
Who is this certification for?
Solutions architects and technical practitioners with hands-on experience building with the Claude Agent SDK, Claude Code, MCP, and the Claude API — the Exam Guide recommends at least 6 months of hands-on practice before attempting the exam.
Related articles
CCAR-F Scenario 6 walkthrough: a JSON Schema enforces shape, never honesty
The last of the 6 official CCAR-F scenarios looks like a tool_use tutorial on the surface. The actual test is whether you know a JSON Schema can force a field to be "string or null" but can never force a model to admit when it's guessing.
CCAR-F Scenario 5 walkthrough: "minimize false positives" is a threshold, not a prompt instruction
The CCAR-F exam guide names the hard part of Claude Code in CI explicitly: minimizing false positives. That's not solved by asking the model to be conservative — it's solved by a confidence threshold that doesn't trust the model's calibration on any single run.
CCAR-F Scenario 4 walkthrough: when a custom MCP tool is justified (and when it isn't)
Scenario 4 of the CCAR-F exam guide mixes built-in tools with MCP servers. The exam-relevant judgment isn't knowing MCP syntax — it's knowing which of the two to reach for, and building a trap codebase makes the difference impossible to fudge.